Mukund Ummadisetti

Software Engineer · AI / Computer Vision

Regents' Scholar studying CS at UC Irvine. Most recently a Software Engineering Intern at Wells Fargo, where I shipped SMS consent management APIs serving millions of users. Currently building computer vision systems for autonomous UAVs at UCI UAV Forge and co-authoring an accepted NASA mission proposal for autonomous proximity operations.

View ProjectsExperience
Mukund Ummadisetti

Experience

Wells Fargo logo

Wells Fargo

Software Engineering Intern · Consumer Technology, Cross-Channel Banker Platform

Jun 2025 – Aug 2025 · Dallas, TX

  • Built SMS consent management APIs serving millions of users with OAuth 2.0 auth and RESTful endpoints in Spring Boot microservices, achieving 98% code coverage.
  • Delivered an end-to-end feature for mobile consent flows: led design discussions with product and platform stakeholders, wrote technical specs, and drove the feature through code review, QA, and production rollout.
  • Maintained and troubleshot services across dev, test, and production environments.
JavaSpring BootREST APIsOAuth 2.0Microservices
PledgeStar logo

PledgeStar

Software Development Engineering Intern

Jun 2021 – Aug 2021 · May 2022 – Jul 2022 · Remote

  • Shipped a cross-platform mobile app (iOS, Android, Windows) with Node.js-backed services, reducing deployment time by 40% and scaling to 10,000+ users with 99.5% uptime.
  • Architected a venue-type abstraction layer in LightWave enabling 5 new venue types; conducted large-scale field testing at 12 venues with zero critical production issues.
DelphiNode.jsAWS EC2Cross-platformMobile

Projects

Autonomy & Perception

UCI UAV Forge: Perception

Improved aerial object detection accuracy by 32% via fine-tuned YOLO and deployed VGGT Transformer for real-time depth mapping.

Computer vision stack for autonomous aerial systems. Built Python data pipelines for synthetic training data generation, fine-tuned YOLO in PyTorch with dataset augmentation and hyperparameter tuning, and deployed a VGGT Transformer for real-time depth mapping and 3D point cloud generation, reducing inference latency 40% via model quantization for deployment on embedded flight hardware.

PythonPyTorchYOLOVGGTComputer Vision
Team site →View on GitHub →

UCI CanSat: Controls Team

Real-time telemetry visualization and STM32 embedded firmware for the CanSat competition.

Built a PyQt5 + Matplotlib dashboard rendering live telemetry charts (altitude, acceleration, gyroscope, voltage, current) with rolling-window updates. Developed embedded C/C++ firmware on STM32 for real-time telemetry, sensor fusion, and flight-state management, with flash-based state persistence for fault recovery during mission-critical operations.

PythonPyQt5MatplotlibC/C++STM32ESP32
Team site →View on GitHub →

NASA L'SPACE: Autonomous Proximity Ops

Co-authored accepted NASA mission proposal for autonomous proximity operations.

Proposal Development Fellow on the NASA L'SPACE NPWEE program. Contributed to mission architecture and autonomy systems design for in-space rendezvous scenarios (TXO 4.1.1 / 4.3.2 / 4.4.3).

AutonomyAerospaceMission DesignSystems Engineering
Program info →

Software Engineering & Machine Learning

LightWave

Cross-platform light-show app: 10,000+ users, 99.5% uptime, field-tested at 12 venues.

Cross-platform app (iOS, Android, Windows) built at PledgeStar. Architected a venue-type abstraction layer enabling 5 new venue types with dynamic themes. Field-tested at 12 venues with zero critical production issues.

DelphiAWS EC2Cross-platformMobile
Watch demo →OC Register article →

Chess Vision Engine

End-to-end chess position recognition: board detection → CNN piece recognition → FEN → Stockfish analysis.

Detects a chessboard from an image, reconstructs the position, and evaluates it with Stockfish. Includes board detection, piece recognition (CNN), FEN generation, a web UI, and engine analysis.

PythonOpenCVCNNStockfishFastAPI
View on GitHub →

Brian: Enterprise Agent Memory Platform

Persistent, queryable memory for enterprise agents, built at LA Hacks 2026.

Autonomous agentic workflows orchestrated with LangGraph (Reader/Writer agents performing multi-step planning, memory retrieval, and tool-use). MCP server exposes retrieval tools to coding agents over a two-stage vector + BM25 RAG pipeline with cross-encoder reranking, deployed via FastAPI on PostgreSQL + pgvector, ingesting GitHub, Slack, and meeting transcripts.

PythonFastAPILangGraphPostgreSQLpgvectorMCPTypeScript
LA Hacks →View on GitHub →

Contact

UC Irvine · B.S. Computer Science, Donald Bren School of ICS